def evaluate_incomplete_entry(classifier_cls):
'''
TODO: Fill out the function to compute marginal probabilities.
Params:
- classifier_cls: either NBClassifier or TANBClassifier
- train_subset: train the classifier on a smaller subset of the training
data
Returns:
- P_c_pred: P(C = 1 | A_observed) as a scalar.
- PA_12_eq_1: P(A_12 = 1 | A_observed) as a scalar.
'''
global A_base, C_base
# train a TANB classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
c_pred, logP_c_pred = classifier.classify(entry)
P_c_pred = np.exp(logP_c_pred)
print(' P(C={}|A_observed) = {:2.4f}'.format(c_pred, P_c_pred))
# TODO: write code to compute this!
a_pred, logP_a_pred = classifier.unobserved_probability(entry, index=11)
PA_12_eq_1 = np.exp(logP_a_pred)
print(' P(A={}|A_observed) = {:2.4f}'.format(12, PA_12_eq_1))
# PA_12_eq_1 = None
return P_c_pred, PA_12_eq_1
def predict_unobserved(classifier_cls, index=11):
global A_base, C_base
# train a classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
a_pred = classifier.predict_unobserved(entry, index)
print(" P(A{}=1|A_observed) = {:2.4f}".format(index + 1, a_pred[1]))
return
def evaluate_incomplete_entry(classifier_cls):
global A_base, C_base
# train a classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
c_pred, logP_c_pred = classifier.classify(entry)
print(" P(C={}|A_observed) = {:2.4f}".format(c_pred, np.exp(logP_c_pred)))
return
def evaluate_missing_posterior(classifier_cls, missing_idx):
global A_base, C_base
# train a TANB classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
m_pred, logP_m_pred = classifier.missing_posterior(entry, missing_idx)
print(' P(M={}|A_observed) = {:2.4f}'.format(m_pred, np.exp(logP_m_pred)))
return
def evaluate_incomplete_entry(classifier_cls):
global A_base, C_base
# train a TANB classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
c_pred, logP_c_pred = classifier.classify(entry)
print ' P(C={}|A_observed) = {:2.4f}'.format(c_pred, np.exp(logP_c_pred))
return
def evaluate_incomplete_entry(classifier_cls):
global A_base, C_base
# train a TANB classifier on the full dataset
classifier = classifier_cls(A_base, C_base)
# load incomplete entry 1
entry = load_incomplete_entry()
prob = classifier.shuffle(entry)
print ' P(C={}|A_observed) = {:2.7f}'.format(1, prob)
return